A Preliminary Examination Of The Impact Of Working Memory Training On Syntax And Processing Speed in Children With ASD Part 4

Results

After verifying that variables met standard assumptions of normality and heterogeneity, we checked for correlations between WM scores and syntax, as well as correlations of WM/syntax to clinical variables (i.e., age, non-verbal reasoning, and autistic symptomatology). As a first step in the analyses, we created composite variables for WM, syntax, and attention measures (see “preliminary analyses” for details of these calculations). 

There is a close relationship between grammar and memory. When we learn grammar, we need to repeat and practice it, which will help us better remember the grammar rules and structures. Mastering the rules of grammar makes our writing more accurate and fluent, thereby improving our ability to express ourselves in communication.

At the same time, learning grammar can also improve our thinking ability and logical thinking ability. In the process of learning grammar, we need to constantly reason and analyze sentence structures, which makes our brains more flexible and agile and can better help us solve problems.

In addition, the study of grammar can also help us improve our memory. Some details and important knowledge points in grammatical rules require us to constantly remember and master them. This memory process can promote continuous stimulation and training of our brains, thereby improving our memory and concentration.

In short, mastering grammar is of great help to our study and life. We need to actively learn and apply grammar, which will not only improve our expression and thinking abilities but also help us improve memory and learning efficiency, laying a solid foundation for future success. It can be seen that we need to improve our memory, and Cistanche deserticola can significantly help us improve our memory because Cistanche deserticola is a traditional Chinese medicinal material that has many unique effects, one of which is to improve memory. The efficacy of minced meat comes from the various active ingredients it contains, including acid, polysaccharides, flavonoids, etc. These ingredients can promote brain health in various ways.

Click know ways to improve brain function

These composites were subjected to repeated-measures ANOVA, to test for the main effect of the intervention (T1, T2, T3) on WM (direct effects) and syntax and attention (transfer effects). Those composite measures that showed significant change over time were further subjected to repeated-measures ANOVAs to compare performance from T1 to T2 (intervention effects) and from T1 to T3 (long-term effects); we also conducted exploratory analyses of individual WM, syntax, and attention measures, to identify specific domains where the intervention had the greatest impact. Finally, additional analyses probed clinical and cognitive predictors of changes in WM and syntactic ability.

Preliminary Analyses

Before exploring the effects of WM training, we ascertained whether WM and syntax were correlated at the T1 baseline, as in previous studies (Durrleman & Delage, 2016; Schuh & Eigsti, 2012). We calculated unit-weighted standardized (Z-score) composite scores for simple and complex spans; each measure contributed equally to the composite. The simple span composite was the average of forward digit recall and serial order word span Z-scores; these individual measures were highly correlated, r(30)=0.60, p<0.001. 

Nonword repetition (phonological WM) was excluded from this composite, as it was uncorrelated with serial order word span, r(30)=0.29, p=0.11. The complex span composite was the average of backward digit recall and counting span z-scores, which were also highly correlated, r(30)=0.47, p=0.006. Results showed that the complex but not simple span composite correlated with all syntactic measures; see Table 4. We also tested correlations of syntax and WM with age, non-verbal reasoning (Raven’s matrices), and autistic symptomatology (CARS score); none were significant. 

The syntax composite was calculated as the average of the Z-scores of (1) Elicited production of root questions, (2) Elicited production of accusative clitics, (3) Sentence repetition, and (4) Comprehension of complement sentences. The attention composite was calculated as the average of the Z-scores of (1) Sky Search and (2) Opposite Worlds tasks.

Improving Working Memory: Direct Effects

To test whether WM training led to significant improvements in WM performance, we performed an initial repeated-measures ANOVA on T1, T2, and T3 scores. Both the simple WM composite scores, F(1, 29)=27.89, p<0.001, Cohen’s f=0.93, and complex WM composite scores, F(1, 28)=185.89, p<0.001, Cohen’s f=2.48, showed significant main effects of training with large effect sizes. As such, we performed follow-up analyses to compare T1 with T2, for each of the WM measures; scores and statistics are presented in Table 5. Results indicated a significant effect of training, with a medium to large effect size, for all WM tasks, with improvements from T1 to T2.

Improving Syntax: Transfer effects

Repeated-measures ANOVAs tested whether training effects transferred to syntax. An initial repeated-measures ANOVA on T1, T2, and T3 composite syntax scores revealed a significant effect of training with a large effect size, F(1, 26)=98.16, p<0.001, Cohen’s f=1.86. As such, we performed follow-up analyses to compare T1 with T2, for each of the syntax assessments; scores and statistics are presented in Table 6. 

Each of the measures showed a mean increase in accuracy, though not all changes were statistically meaningful. Of the four elicited syntactic production measures, root questions significantly improved between T1 and T2; of these root questions, only object questions, including wh-in situ items, showed significant change, t(29) = − 2.2, p = 0.03, d = 0.4. 

There was no significant change in wh-fronted questions or object clitic productions. For complex sentence repetition, the number of correctly repeated syllables significantly improved from T1 to T2, with a medium-to-large effect. This result cannot only be explained by better memory because repetition accuracy for simple sentences (which contained an identical number of syllables) showed no significant improvement, t(29)=− 1.2, p=0.2. The percentage of sentences in which the degree of embedding was respected, independently of the lexical units, also improved. Finally, the children also showed a significant improvement in complement sentence comprehension between T1 and T2.

Improving Attentional Skills: Transfer Effects

To test the hypothesis that WM training would boost selective attention, processing speed, and attention shifting, we conducted a third repeated-measures ANOVA on the composite T1, T2, and T3 attention scores; this revealed a significant effect of training, with a large effect size, F(1, 25) = 442.31, p < 0.001, Cohen’s f = 4.04. As such, we employed a series of repeated-measures ANOVAs, detailed in Table 6, to test specific effects. All measures showed a significant decrease in reaction time between T1 and T2, revealing faster processing. The largest effect was observed for the most demanding task, the Opposite World task, which required the participant to inhibit a prepotent response.

Long‑Term Effects

Of the original sample of 30, 26 children with ASD were retested at T3, three months after the posttest. Repeated measures ANOVAs with time (T1, T2, T3) as a within-subjects factor revealed significant main effects of time were observed for each of the WM tasks; see Table 5. Post-hoc Tukey’s HSD tests explored the periods for which significant change occurred. Performance at T3 was significantly better than T1 for the three simple-span tasks (forward digit span, nonword repetition, and serial order word span), but not for the two complex-span tasks. Comparisons of T2 and T3 showed no significant changes except in backward digit recall which showed a significant decrease (e.g., worse performance). Figure 2 illustrates these results for the serial order word span (2a) and forward digit recall (2b), which followed the predicted pattern of significant improvements between T1 and T2 and between T1 and T3 and no change between T2 and T3. Performance on the backward digit recall task (2b), however, decreased significantly from T2 to T3.

The long-term stability of transfer effects was calculated for those measures of syntax and attention for which a significant T1-T2 increase was observed. As indicated in Table 6, the gains in syntactic abilities observed at T2 were still present at T3, with no significant decrement in performance. This pattern held for both syntactic measures (elicited production of root questions, sentence repetition, see Fig. 2c) and for each of the three attentional tasks (2d).

Further Analyses

The results revealed a significant impact of WM training on WM itself (direct effects) and attention and syntax (transfer effects); improvements in the latter were in the moderate range. Given the heterogeneous nature of ASD, it was important to test whether improvements were observed for all participants. We calculated gains for the primary measures4 by subtracting the T2 results from the T1 results for each child. Appendix E presents these measures of gains for each child as well as group means for each measure. As expected, improvements were extremely variable with some children making progress on all the tasks (such as participants 4 or 30) and others showing no real improvement on any task (such as participants 17 or 20). We tested whether age, non-verbal reasoning, or autistic symptomatology predicted gains, using correlational analyses; none were significant, and all p’s>0.10.

Discussion

Our study explored the effects of an intensive WM training program, Magic Memory (Delage et al., 2017), on WM, syntax, and attention, for 30 French-speaking children with ASD aged 5;11 to 11;10. We expected similar results in ASD as those previously reported for similarly-aged children with DLD, namely, direct effects on WM as well as transfer effects on expressive syntax (Delage et al., 2020, 2021; Stanford et al., 2019). In the present work, we aimed to replicate and expand upon our previous studies with participants with DLD, and related studies reporting WM/syntax links in ASD (Durrleman & Delage, 2016; Riches et al., 2010; Schuh & Eigsti, 2012; Weismer et al., 2017). The current study employed measures used in previous training studies in DLD, along with several additional tasks:

We predicted improved performance on the capacities directly trained, i.e. on WM, as well as on syntactic and attention domains not directly trained but hypothesized to be related to WM. We also predicted that these gains would be maintained three months after training.

Links Between WM and Complex Syntax

Preliminary analyses confirmed the close relation between complex-span measures (i.e., complex WM) and all measures of syntax in our participants with ASD, replicating previous results (Durrleman & Delage, 2016; Riches et al., 2010; Weismer et al., 2017). The fact that simple spans did not appear to be linked to the same extent as syntactic capacities in our population suggests that the more executive component of WM plays a role in complex syntactic processing in ASD. This result echoes the findings of Delage and Frauenfelder (2019) who reported that complex spans (but not simple ones) predicted measures of syntactic complexity in spontaneous language samples of 48 TD children; this pattern was reversed in DLD (Delage & Frauenfelder, 2020). 

This larger pattern of results suggests that the WM deficits in DLD reflect reduced phonological storage (see also Alt, 2011; Gathercole & Baddeley, 1990), whereas the WM deficits in ASD reflect executive dysfunction, notably in mental flexibility (Demetriou et al., 2018). The current results are consistent with this hypothesis, as our participants with ASD had significantly better simple-span results (mean Z-score of − 1.2) relative to complex-span ones (mean Z-score of − 2.5). Their difficulties in processing complex sentences likely reflect the inefficient or otherwise more impaired performance of the cognitive operations required for complex spans (i.e., reduced cognitive flexibility to cope with interference during verbal storage).

Direct Effects of WM Training

The WM training was effective as it led to significant improvements in performance on all WM tasks. Admittedly, these tasks closely resembled the activities presented in the Magic Memory program, with distinct visual and verbal content and a distinct format (paper versus computerized). Nevertheless, such direct effects have not been consistently observed (Majerus, 2016; Melby-Lervag & Hulme, 2013) and it was encouraging to see that the children were able to transfer their skills from one format to another.

Effect sizes were impressive, with medium to large effects, except non-word repetition, for which improvements were significant with a small effect size. Speech-sound difficulties have been reported for subgroups of children with ASD (Kjelgaard & Tager-Flusberg, 2001; Wolk et al., 2016; Zebib et al., 2013), and such difficulties may have obscured the training effects. The serial-order word-span task, in contrast, displayed the highest effect size. This task of putting animal cards on a podium in order is very similar to the training program’s task of putting pictures corresponding to familiar words into train cars in order, a similarity that undoubtedly contributed to the large gains that participants displayed in this particular task. 

In addition, this serial memory task requires participants to remember the order of animals participating in the race, without necessarily retaining phonological representations of the animals’ names (Majerus, 2008; Majerus et al., 2006). Thus, this task focuses on WM while minimizing the influence of language, and of potentially degraded phonological representations. The fact that the task that least depends on phonological representations showed the most pronounced training effects is consistent with the aims of the training, which is meant to enhance “pure” memory processes (while also utilizing verbal material). These results suggest that this program will be effective for other conditions characterized by phonological deficits, including some forms of ASD and children with speech-sound disorders (Claessen & Leitão, 2012; Gathercole & Baddeley, 1990; Leonard, 2014; Zebib et al., 2013).

Transfer Effects of WM Training on Syntax

All syntactic measures, both expressive and receptive, showed a mean increase in accuracy, with significant improvement in the production of root questions, and in the repetition and comprehension of complex sentences. Such transfer effects cannot be attributed to the material used in training, since the WM program presented only isolated words. Results of the sentence repetition in particular were striking, because they suggested significant increases at post-test for complex but not for simple sentences, though both types of sentences were matched in length. Similarly, independent of sentence length, participants showed improvement in producing complex sentential embedding.

Taken together, these exciting results suggest a meaningful transfer of WM improvement into the domain of complex expressive syntax, extending previous findings of training effects for children with DLD (Delage et al., 2021). Not all of the previous DLD results were replicated here; for example, there was no significant improvement in the production of accusative clitics (Stanford et al., 2019). Differences between response patterns in DLD and ASD likely reflect the nature of the tasks and the deficits specific to the autistic spectrum. For example, the sentence repetition task involves limited social interaction; participants simply repeat the target sentence. 

In contrast, the clitic production task, which is understood by age four in typical development (Delage et al., 2016), requires more direct interaction with the experimenter, who asks questions (ex: Look! What is the man doing with his car? Tell me!) about pictured scenes. While the social demands are highly structured and relatively small, participants must listen and respond to the examiners; this could be a more difficult task for children with ASD (DSM-5, APA, 2013). However, the current results were not consistent with this hypothesis, as there was no relationship between syntax scores and autistic symptomatology (CARS scores). A more detailed assessment of the social communication skills, using the Children’s Communication Checklist (CCC-2, Bishop, 2003) for example, might have been more sensitive to individual differences in pragmatics.

Results also showed significant improvements in the comprehension of complement sentences. Receptive skills were not assessed in our training studies of children with DLD, who do display deficits in receptive syntax (Delage & Frauenfelder, 2020; Friedman & Novrogrodsky, 2004). Such receptive difficulties are widely reported in children with ASD, including deficits in the comprehension of complex wh-questions and relative clauses (Durrleman et al., 2016), passives (Durrleman et al., 2017), clitics (Terzi et al., 2014), and more globally on standardized measures of receptive syntax and morphology (Brynskov et al., 2017). Difficulties in comprehension of object relatives have been found to persist in young adults (Durrleman et al., 2015). In 2019, Durrleman et al. showed that comprehension of complement sentences can be improved in children with ASD, aged 5 to 11, through a brief training targeting sentential complements. In this study, WM training provided benefits in both cognitive and linguistic domains. Future work could compare the effects of training on complex syntax (as in Durrleman et al., 2019) to the effects of WM training, to test whether the effects are bidirectional. Such studies will illuminate language and cognitive deficits in this population.

Transfer Effects of WM Training on Attention

In addition to the direct effect on WM capacities, WM training had an indirect effect on other attentional abilities, which is unsurprising given their theoretical overlap (Baddeley, 2003; Barrouillet & Camos, 2012; Engle, 2002; Majerus et al., 2009; Veer et al., 2017). Indeed, participants showed improved speed with no concomitant decrease in accuracy for the three aspects of attention: selective attention, processing speed, and attention shifting. These gains suggest that WM training provides a cognitive “boost,” allowing children to process information more quickly, thereby freeing up processing resources. This change likely has a snowballing or cascading effect on syntactic processing as the reduction of cognitive limitations would influence the processing of computationally complex structures, such as those including embedding and/or a movement operation (Delage & Frauenfelder, 2019, 2020; Jakubowicz & Strik, 2008; Tuller et al., 2012). Attention limitations are also considered by Chomsky (2005) to impact the processing of syntax, and attention is seen as underlying the development of executive functions, such as WM and inhibition (Garon et al., 2008); given this foundation, we hypothesized that training in WM could lead to improvements on both attentional and syntactic measures.

Although previous WM training studies conducted with children with DLD have found the same direct and indirect effects (on syntax), these studies did not assess attentional abilities and were therefore unable to test for the transfer effects on attentional tasks that we obtained in the current study. Future studies should also evaluate these transfer effects in children with DLD. Similar effects would be expected, consistent with studies linking attentional resources to the processing of complex syntax in children with DLD (Montgomery et al., 2009; Stanford & Delage, 2020). Moreover, it would be informative to contrast the effects of pure training of the attentional component (such as the TALI, Kirk, et al., 2016) with the effects in the current study, and to compare the effects of both on syntax; this would reveal whether the linguistic gains following intensive WM training can be attributed to WM improvements or better functioning of the attentional system.

Follow‑Up

Almost all participants (26/30) were retested three months after training to assess long-term effects. Results showed that improvements were maintained in all WM tasks except backward digit recall, for which performance decreased from T2 to T3 (although T3 performance was still significantly better than at baseline). This task requires reversing the order of previously heard stimuli (digits). The training program was similar, but involving color names rather than digits. It is possible that the extensive overlap between tasks boosted the progress observed between T1 and T2, and that these gains were not maintained without practice (see Fig. 2b). Aside from this task, the overall results point in the direction of an encouragingly persistent and long-term improvement in WM performance, in contrast to previous WM training studies with healthy adults (Melby-Lervag & Hulme, 2013; Melby-Lervag et al., 2016). This suggests that WM training might be most effective with children, whose development is ongoing, or for individuals with specific WM defects.

Long-term transfer effects were similarly positive, with improvements maintained on all measures showing T1 to T2 gains, whether in the areas of syntax or attentional skills. This result is particularly promising since we were unable to demonstrate the maintenance of syntactic transfer effects in previous work, likely because only a small subset of children with DLD were retested (12 out of 32; Delage et al., 2021).

Limitations

The present study did not include an active control group, which would have conclusively demonstrated that, without specific training, the performance in WM, syntax, and attention would show significantly fewer improvements. As discussed above, we chose not to include such a group because our previous studies of children with DLD or TD revealed no benefit to WM or syntax from training focusing on academic skills (Delage et al., 2021; Stanford et al., 2019). While this is a significant limitation, we argue that the T2-T3 comparison provides a type of control, in line with ABAB treatment paradigms (e.g., Kratochwill et al., 2013). Such studies involve alternation between a baseline period (A) and a treatment period (B), sometimes including only a single subject. In such paradigms, the specific effects of the program should be visible following the training periods (B) and more discrete following the non-active (baseline) periods (A). This was precisely the pattern observed in the current study, in which the children’s performance did not improve between T2 and T3, suggesting that the progress observed from T1 to T2 was a function of the training program, rather than general maturation and development.

We also acknowledge that transfer effects from WM training onto syntax could be explained by other reasons than attentional resources and computational processing. Indeed, training sessions involve interactions between the child and the caregiver, as well as the experimenter. The increase in conversations with this diversity of interlocutors may also have helped to motivate children with ASD to engage in language tasks. To explore this aspect, it would have been useful to administer a questionnaire on the children’s social interactions, such as the Social Communication Questionnaire (Rutter et al., 2003a, 2003b), before and after the training. We leave this for future work. Other limitations stem from the fact that we did not evaluate the capacities of children to generalize their improved syntactic skills to other daily life contexts, which would have been possible if we had carried out an analysis of spontaneous language. Complex syntax, assessed by such an analysis of spontaneous language samples, has already been shown to be strongly linked to WM in children with TD and DLD (Delage & Frauenfelder, 2019, 2020). We would thus expect that our training would also improve the children’s spontaneous syntax, yielding richer productions in conversational contexts. Moreover, it should be noted that our protocol was the same for all children, whereas the symptoms/traits of ASD are known to be heterogeneous. Adapting the material and the setting to the particularities of each child would certainly be useful in clinical practice, but this would not be suitable for a rigorous experimental design.

We should finally note that our statistical approach involved starting with an omnibus test of change in broad cognitive domains, with significant results to be followed by detailed exploratory tests of change in specific local task domains. While our results are compelling and seem to be highly consistent across within-domain measures, this approach requires further replication given the relatively small sample size and the number of comparisons.

Conclusion

Following WM training, our participants showed improvements not only in WM but also in distal domains such as speed of attentional processing and expressive and receptive syntax. Of course, these changes might reflect a general g-factor, such that the children with stronger cognitive abilities pay more attention in the training and show the greatest improvements. To address this possibility, future work could include a dynamic measure of cognitive functioning to directly examine the child’s response to learning (Camilleri & Law, 2007). Dynamic assessment of learning capacity provides a reliable and valid measure of general intellectual functioning and is a good predictor of future learning (Hessels & Hessels-Schlatter, 2010). Moreover, this type of assessment would be highly appropriate for children with emotional or personality challenges that could interfere with their performance (Tzuriel, 2001). A dynamic assessment could illuminate the inter-subject variability we observed in this study, with ASD participants who benefit more from WM training being those who show a better capacity to learn.

While they must be replicated, these exciting results provide the impetus for further studies of WM interventions. Thereafter, the training materials should be available to children with WM or syntactic impairments, regardless of the proximal source of those impairments (e.g., DLD or ASD).

The findings of this study provide further evidence in support of the efficacy of WM training in multiple domains and fall within an ‘evidence-based practice’ framework, which emphasizes the role of research findings in clinical decision-making (Sackett et al., 2000).

Author Contributions

All authors contributed to the study's conception and design. Material preparation, data collection, and analysis were performed by Hélène Delage, Inge-Marie Eigsti, and Emily Stanford. The first draft of the manuscript was written by Hélène Delage and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Declarations

Conflict of interest There are no real or potential conflicts of interest related to the manuscript.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution, and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third-party material in this article are included in the article's Creative Commons license unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license.

References

1. Adams, A. M., & Gathercole, S. E. (2000). Limitations in working memory: Implications for language development. International Journal of Language and Communication Disorders, 35(1), 95–117.

2. Alloway, T. P. (2018). Working memory and clinical developmental disorders: Theories, debates, and interventions. Routledge.

3. Alloway, T. P., Rajendran, G., & Archibald, L. M. D. (2009). Working memory in children with developmental disorders. Journal of Learning Disabilities, 42, 372–382.

4. Alloway, T. P., Seed, T., & Tewolde, F. (2016). An investigation of cognitive overlap in working memory profiles in children with developmental disorders. International Journal of Educational Research, 75, 1–6.

5. Alt, M. (2011). Phonological working memory impairments in children with specific language impairment: Where does the problem lie? Journal of Communication Disorders, 44(2), 173–185.

6. American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). American Psychiatric Association Publishing.

7.Baddeley, A. D. (2003). Working memory and language: An overview. Journal of Communication Disorders, 36(3), 189–208.

8. Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a “theory of mind?” Cognition, 21(1), 37–46.

9.Barouillet, P., & Camos, V. (2007). Le développement de la mémoire de travail. In J. Lautrey (Ed.), Psychologie du développement et de l’éducation (pp. 51–86). PUF.

10.Barrouillet, P., Bernardin, S., & Camos, V. (2004). Time constraints and resource sharing in adults’ working memory spans. Journal of Experimental Psychology: General, 133(1), 83.

For more information:1950477648nn@gamil.com

Alloway, T. P., Rajendran, G., & Archibald, L. M. D. (2009). Working memory in children with developmental disorders. Journal of Learning Disabilities, 42, 372–382.